[Arm-netbook] EOMA68-A20, MD 1.7 severe Power Problems

[Luke Kenneth Casson Leighton]

On Sun, Jul 26, 2020 at 10:33 AM Pablo Rath <pablo at parobalth.org> wrote:
>
> I am very sorry to inform everyone on this list that I had a severe power problem with my Micro Desktop 1.7.
> I applied power to the DC Jack and the area right to the jack burned
> out. (see attached picture).

thank you for sending this to the list, as i asked, after you sent it
initially privately.  i had this happen to a 1.5 MD board 3 years ago,
but no others, despite them running for prolonged periods of time.

what i noticed about that board was that the inductor was not properly
soldered down.  this would be insufficient contact, introduce
resistance, and at that point the RT8288 would go unstable.


> One thing that I noticed today was that UART worked when the
> card was powered up whilst attached to USB OTG alone (DC
> Jack on MD not connected)!

yes, i mentioned this already.  the TX and RX line GPIO current is
sufficient to "power" the LEDs and possibly the FT2322 IC as well.

> But according to the wiki this
> should not work at all because of the Y6280 current control
> IC set at 1A acting as a diode in this case.

no: it's not being powered through the 5.0V rail: the LEDs on the
USB-UART are being powered through the *processor*, through the TX and
RX lines.

> When I applied power only to the MD I noticed that UART
> output was complete garbage (random characters).

yes.  i did say.  you get GND loops that spike the USB-UART
sufficiently to trigger the RX line.

> I don't know if only
> sometimes or everytime. When on DC power, board in FEL-mode with no card
> and pressing '2', then connecting USB OTG and booting U-Boot sunxi over FEL everything worked flawlessly.
>
> As I said I am very sorry that I screwed up.
> Luke, do you have any ideas what went wrong?

not in the slightest.  or - maybe: have a look at the contact points
where the inductor sits on the PCB.  there should be quite a lot of
solder, there.


> Do you think this
> +also destroyed the computer card?

when it happened for me it did no damage.  all i did was (because i
didn't have any spares) find a USB2 back-to-back power cable (i may
have made one by cutting a plug off a USB device and wiring it to a
5.0v supply), plug it into one of the USB2 ports and provided "direct"
5.0v power that way.  you *need* a stable supply to do that (1.5
preferably 2.0 A) designed *specifically* for providing USB power.
UNDER NO CIRCUMSTANCES plug the 12v PSU into the USB socket.  and DO
NOT use an "off-the-shelf generic 5.0v wall wart".  use something
SPECIFICALLY designed for providing USB power because it is (a) stable
and (b) current-limited.

if you plug the Card directly into a socket (OTG, removed from the
MicroDesktop) - not via a USB hub - "ls" should show the familiar USB
ID for the A20.


> If the error is on my side (embarassing but still the better
> +case) I
> will have to swallow the bitter pill that I screwed up badly.
> I am still going to cover the costs of computer card and MD so there will not be a financial loss to the project.

appreciated


> I am very disappointed as I have made could progress. Had sunxi U-Boot,
> sunxi 3.4.104 kernel running and booting Debian Stretch rootfs...

great!

> On the other hand if this is a general problem
> better we know it now and can take the necessary steps.

modifying the 1.7 MD PCB and going through *yet another* round of PCB
development costs and time delays - this not something i want everyone
to have to go through.   particularly because they're already
manufactured.

sigh.

l.